Background Despite growing evidence showing an association between Parkinson's disease (PD) and diabetes, epidemiological studies have shown conflicting results. Aims of the study To evaluate the association between PD and diabetes and the impact of diabetes duration in this association in an elderly (≥65 years) Spanish population. Methods Data for this cross‐sectional population‐based analysis were obtained from NEDICES study. Subjects were identified from census list. Diagnosis of PD was confirmed by neurological examination. Diabetes was defined by self‐report, being on antidiabetic medication or diagnosis on medical records. Logistic regression analysis adjusted by potential confounders was performed to estimate the association between both conditions and also after dividing patients into short‐duration (<10 years) and long‐duration (≥10 years) diabetes. Results A total of 4998 subjects were included (79 PD and 4919 controls). Univariate analysis did not show any association between prevalence of PD and diabetes (OR 1.89, 95% CI 0.90‐3.98, P=.09), although subgroup analysis showed a positive association in those with long‐duration diabetes (3.27, 95% CI 1.21‐8.85, P=.02). Conclusions Diabetes duration might be an important factor in the association between PD and diabetes, and the risk might be limited to those with longer disease duration.